HELENA:Hierarchical Sparse Coordination over a Union of Complementary Topologies for MAS
3.40T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2608.04634.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryHELENA is a multi-agent LLM framework that builds a union of complementary topologies selected via Monte Carlo Tree Search and Determinantal Point Process, then activates only a sparse subgraph per step while agents exchange compressed latent briefs, followed by local self-refinement. Reports SOTA on eight benchmarks with an average 3.47% gain over the strongest baseline.
Why it mattersA concrete multi-agent topology framework with measured benchmark gains; useful for practitioners building agent orchestration beyond a single graph.
Cited by
No citations on record.